Transferring Automotive Optimization Methods from Static to Dynamic Function Distributions

In the fast-changing automotive industry, optimizing vehicle architectures has become more complex and crucial. Traditional static architectures, where software components (SWCs) are fixed to specific electronic control units (ECUs) during production, can no longer meet the dynamic demands of modern vehicles. This paper addresses these limitations by introducing a novel approach for transferring automotive optimization methods to support dynamic function distributions. Thereby we enable the shifting of SWCs across various execution nodes at runtime, including high-performance controllers (HPCs), cloud, and edge nodes, based on current operational requirements and environmental conditions. We explore the underlying principles of dynamic function distribution, identify key use-cases, and analyze the differences between static and dynamic environments to design a generic approach to dynamic optimization. A prototypical implementation is presented and validated through illustrative examples, demonstrating the feasibility and benefits of this approach with up to approximately 50% of onboard power savings. Our findings highlight the potential improvements in vehicle performance, energy efficiency, and flexibility.